This series of files compile all analyses done during Chapter 2:

All analyses have been done with R 3.6.2.

Click on the table of contents in the left margin to assess a specific analysis.
Click on a figure to zoom it

To assess Section 2, click here.
To go back to the summary page, click here.


Sources of activity considered for the analyses:

Fisheries data considered for the analyses (expressed as number of fishing events or kilograms of collected individuals for each gear):

Gear Code Years Events Species
Trap FishTrap 6 (2010-15) 1061 Buccinum sp., Cancer irroratus, Chionoecetes opilio, Homarus americanus
Bottom-trawl FishTraw 2 (2013-14) 2 Pandalus borealis
Net FishNet 1 (2010) 5 Clupea harengus, Gadus morhua
Dredge FishDred 5 (2010-14) 21 Mactromeris polynyma

1. Maps

1.1. General map

1.2. Parameters maps

Depth

Isobaths

Slope

2. Modelling the influence of human activities

We modelled an index of influence for each human activity in order to be used in community distribution models (see section 2). The influence was computed by using an index of exposure and a weighting parameter, because each activity will not have the same relative impact on the ecosystems:

\[ I_{ij} = E_{ij} . w_{j} \]

  • \(I_{ij}\) is the influence index
  • \(E_{ij}\) is the exposure index
  • \(w_{j}\) is the weighting parameter
  • \(i\) is a station
  • \(j\) is a human activity

Two categories of exposure index were calculated seperately: one for land- and sea-based activities and one for fisheries. These indices are relative, with a variation between 0 (low exposure/influence) to 1 (high exposure/influence)

2.1. Calculating the index of exposure

This corresponds to \(E_{ij}\) in Formula 1.

2.1.1. Land- and sea-based activities

The following map present the sources of land- and sea-based human activities.

Here, \(E_{ij}\) have been calculated with connectivity models of particle dispersion.

Methodology

We considered that theoretical particles, which are the resultant of a human activity in the ecosystem (e.g contaminants, sediment), can difuse in the area from the source of the activity with different physical constraints. The general function to calculate \(E_{ij}\) is:

\[ E_{ij} = f_{j} \left( D_{ij}, C_{i} \right) \]

  • \(D_{ij}\) is the distance from the source
  • \(C_{i}\) are the physical constraints
  • \(f_{j}\) is the connectivity function
  • \(i\) is a station
  • \(j\) is a human activity

This step used methods from the package gdistance with which we established a connectivity matrix to calculate the trajectory of the particles released from the source of an activity. This matrix is based on a “resistance seascape” concept, where each cell of the raster has a cost to be included in the path between the start (source of the activity) and end (station) points. The final path is computed with a least-cost path algorithm, and the higher the distance is, the lower the influence.

To calculate the cost of each cell for the connectivity matrix, we considered physical constraints to account for natural relationships. We considered three underlying principles:

  • marine ecosystems: particles cannot disperse on land
  • gravity: particles disperse easily from shallow to deeper depths, while the reverse will be difficult
  • hydrodynamism: particles disperse according to local hydrodynamical currents

The dispersion model then considered: (i) coasts and islands as boundaries unselectable by the least-cost algorithm, (ii) bathymetry, (iii) river plumes as hydrodynamical fronts with an intensity and a direction (a complete circulation model in BSI is not yet available).

If we need to include a specific behaviour depending on the distance travelled by the particle, different “decay” functions may be used to account for dispersion. Here, we used a linear relationship as a first iteration, so that a particle will disperse the same way when close or far from its source.

Because each activity possess its own influence, we cannot use an identical model for each. Thus, we defined different types of theoretical particles based on their size: small, medium and large. Each type of particle will have a specific connectivity matrix with the same constraints but a different transition function:

Particle Transition function
Small ifelse(d[1] > d[2], 1, 0.9)
Medium ifelse(d[1] > d[2], 1, 0.5)
Large ifelse(d[1] > d[2], 1, 0.1)

The influence of each activity will be a combination of the three types of particle, with three coefficients representing the relative proportions. \(E_{ij}\) is then calculated as the sum of these components:

\[ E_{ij} = \sum_{k}^{s,m,l} f_{p} \left( D_{ij}, C_{i} \right) . p_{jk} \]

  • \(D_{ij}\) is the distance from the source
  • \(C_{i}\) are the physical constraints
  • \(f_{j}\) is the connectivity function
  • \(p_{jk}\) is the proportion of the particle type
  • \(i\) is a station
  • \(j\) is a human activity
  • \(k\) is the particle type (small, medium, large)

Proportions of each particle type will be defined by us, which will necessitate some literature searchs to groundtruth our methodology. Here are the proportions of the small, medium and large particles for each activity:

  small medium large
CityInf 0.6 0.3 0.1
InduInf 0.7 0.2 0.1
DredColl 0.1 0.3 0.6
DredDump 0.1 0.3 0.6
MoorSite 0.4 0.4 0.2
RainSew 0.5 0.4 0.1
WastSew 0.4 0.5 0.1
CityWha 0.4 0.4 0.2
InduWha 0.4 0.4 0.2

Results

The following maps present the values of \(E_{ij}\) for land- and sea-based activities (grey = low exposure; dark blue = high exposure).

CityInf

InduInf

DredColl

DredDump

MoorSite

RainSew

WastSew

CityWha

InduWha

2.1.2. Fisheries

These data belong to Department of Fisheries and Oceans Canada, with a permission granted to David Beauchesne, we cannot present the raw products and we will work on derived data.

Here, \(E_{ij}\) have been calculated with a proxy based on fisheries data for each gear used in the area.

Methodology

We extracted data from a global database for the St. Lawrence, for all fishing events occuring within the Baie des Sept-Îles. Four types of gears (traps, bottom-trawls, nets and dredges) have been used in the bay between 2010 and 2015, to collect eight species (see table at the top of this page).

As each gear was not used consistently during this period, we averaged the number of fishing events to obtain a proxy of fishing intensity. Furthermore, we modified this proxy with a smoothing function in order to ‘diffuse’ the signal around the actual event and to account for incertainty.

Results

The following maps present the values of \(E_{ij}\) for fisheries (grey = low exposure; dark blue = high exposure).

FishTrap

FishTraw

FishNet

FishDred

2.2. Setting the weighting parameters

This corresponds to \(w_{j}\) in Formula 1.

This table shows the weights \(w_{j}\) for each non-fishery human activity:

CityInf InduInf DredColl DredDump MoorSite RainSew WastSew CityWha InduWha
1 1 1 1 1 1 1 1 1

And this one shows the weights \(w_{j}\) for each fishery:

FishTrap FishTraw FishNet FishDred
1 1 1 1

2.3. Calculation of the index of cumulative influence

This corresponds to \(I_{ij}\) in Formula 1.

We can combine each \(I_{ij}\) thanks to the previous calculations through a sum to obtain a cumulative influence index:

\[ CI_{i} = \sum_{j} {I_{ij}} \]

In future iterations, we will try to use different link functions to account for non-additive effects. This score varies between 0 and 13 (number of considered human activities).

Raw index

Classed index

The cumulative influence index has been represented in five classes, according to the colour code of the Marine Strategy Framework Directive (indigo = low influence, less than 20 %; crimson = high influence, higher than 80 %).

Classed index with sources

The cumulative influence index has been represented in five classes, according to the colour code of the Marine Strategy Framework Directive (indigo = low influence, less than 20 %; crimson = high influence, higher than 80 %).

Here are the scores for each sampled station:

These scores will be used for the HMSC models (see section 2).